基于局部拓扑结构的无线传感器网络定位算法研究

发布时间：2019-03-08 14:29

【摘要】：无线传感器网络涉及无线通信、传感器技术、分布式信息处理、嵌入式技术以及微电子等众多技术,在交通、军事、医疗、环保等领域有着广泛的应用。在无线传感器网络众多用途中,确定事件发生的位置是监测到事件发生后需要解决的关键问题之一。位置信息除了确定事件发生的地点外,还具有网络管理、移动目标跟踪、辅助路由等功能。因此设计高效的WSN定位算法是无线传感器网络管理中不可或缺的一部分。本文针对无线传感器网络的定位算法展开了研究。主要的研究工作有:(1)对WSN的定位算法进行了深入的研究,分别从机器学习、测距和非测距三个方面分析总结了WSN定位技术的优势和不足,为设计出高精度、低能耗的WSN定位算法提供了有力的基础。(2)在对基于LE-LPCCA定位算法研究的基础上,引入局部拓扑结构和分布式特性,提出了一种基于局部保持的分布式定位算法LE-DLPCCA。通过仿真实验验证,当训练样本比例为70%,定位精度可达86%,且能耗大幅降低,可延长整个无线传感器网络的生命周期,建模速度同时也提高了8倍。(3)分析了无线传感器网络的拓扑基本符合流形的特点,引入局部拓扑结构和非信标节点的信息,采用半监督学习技术研究节点的定位问题,提出了一种基于拉普拉斯映射的移动节点定位算法LP-LapRLS。该算法不仅提高了映射模型的泛化能力,而且在典型的的流形学习算法中建模效率较高。实验结果表明与同类算法相比LP-LapRLS具有更高的建模效率和定位精度,当训练集比例为60%,定位精度可达84%。(4)在研究了无线传感器网络的体系结构和协议栈的基础上,在VS2010集成环境中采用VC++,设计并实现了WSN定位仿真平台。在该平台中实现了LE-DLPCCA算法和LP-LapRLS算法,最后对这两种基于机器学习的定位算法的定位效果进行了对比与分析。LE-DLPCCA算法相比LP-LapRLS算法定位精度更高,大约提升了2个百分点。但是LP-LapRLS算法在出现离群值的情况下,鲁棒性较好,并且在定位算法中建模效率最高。
[Abstract]:Wireless sensor network (WSN) involves many technologies such as wireless communication sensor technology distributed information processing embedded technology and microelectronics and so on. It is widely used in the fields of transportation military medical protection and so on. In many applications of wireless sensor networks, determining the location of events is one of the key issues to be solved after monitoring the occurrence of events. Location information not only determines the location of the event, but also has the functions of network management, moving target tracking, auxiliary routing and so on. Therefore, the design of efficient WSN positioning algorithm is an indispensable part of wireless sensor network management. In this paper, the localization algorithm of wireless sensor networks is studied. The main research work is as follows: (1) the localization algorithm of WSN is studied in depth, and the advantages and disadvantages of WSN positioning technology are analyzed and summarized from three aspects: machine learning, ranging and non-ranging, so as to design high precision. Low energy consumption WSN localization algorithm provides a powerful foundation. (2) based on the research of LE-LPCCA-based localization algorithm, the local topology and distributed characteristics are introduced, and a distributed localization algorithm LE-DLPCCA. based on local preservation is proposed. The simulation results show that when the proportion of training samples is 70%, the positioning accuracy can reach 86%, and the energy consumption can be greatly reduced, thus prolonging the whole life cycle of wireless sensor networks. At the same time, the modeling speed is improved by 8 times. (3) the topology of wireless sensor networks is analyzed, the local topology and the information of non-beacon nodes are introduced, and the semi-supervised learning technology is used to study the localization problem of wireless sensor networks. A mobile node location algorithm LP-LapRLS. based on Laplacian mapping is proposed in this paper. This algorithm not only improves the generalization ability of the mapping model, but also has high modeling efficiency in the typical manifold learning algorithm. Experimental results show that LP-LapRLS has higher modeling efficiency and positioning accuracy than similar algorithms, when the ratio of training sets is 60%. The positioning accuracy can reach 84%. (4) on the basis of studying the architecture and protocol stack of wireless sensor network, the WSN positioning simulation platform is designed and implemented by using VC in VS2010 integrated environment. In this platform, the LE-DLPCCA algorithm and the LP-LapRLS algorithm are implemented. Finally, the localization effect of the two localization algorithms based on machine learning is compared and analyzed. The LE-DLPCCA algorithm is more accurate than the LP-LapRLS algorithm, and the location accuracy of the two algorithms is higher than that of the LP-LapRLS algorithm. It has increased by about 2 percentage points. However, in the case of outliers, the LP-LapRLS algorithm is robust, and the modeling efficiency is the highest in the localization algorithm.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9;TN929.5

【参考文献】